AI is no longer an abstract acronym on the roadmap – it’s now what most retailers’ transformation programmes are built around.
At the recent Gartner IT Symposium in Barcelona the volume of AI conversation was unmistakable: C-suites have dedicated budgets, transformation timelines are set, and companies are actively layering AI into strategic change programmes.
But while the buzz is real, there is broader talk of an AI bubble which can’t be ignored. Agentic AI undoubtedly delivers key efficiency savings and customer benefits, and the success of widespread implementation will hinge on how it is applied to clear business problems rather than adopted through fear or falling behind.
Why agentic AI matters for retail
Agentic AI – systems which can make decisions and take actions to achieve goals, functioning with a high degree of independence – offers a fundamentally different scale and quality of customer engagement.
Where traditional automation handles scripted transactions, agentic AI can orchestrate multi-step, context-aware conversations across channels, take actions (book, refund, order), and learn from outcomes. The commercial impacts are clear:
- Operational efficiency: AI agents can automate high-volume, low (and increasingly high)-complexity interactions 24/7, reducing pressure on contact centres and lowering labour costs.
- Hyper-personalisation at scale: By combining real-time signals with historical customer data, agents can deliver one-to-one offers, recommendations and recovery journeys that increase basket sizes and conversion rates.
- True omnichannel conversations: Agentic systems can maintain coherent, persistent conversations across OTT apps, RCS, SMS, voice and web chat – letting customers start on WhatsApp and finish in-app or in-store seamlessly.
- Enhanced security and verification: Agentic AI can automate identity checks (silent mobile verification, data checks, behavioural signals) and flag anomalies, reducing fraud while maintaining smooth CX.
These benefits are compelling, but implementation is complex.
Reliable agentic AI demands integrated data, robust guardrails, audit trails, multilingual support, seamless human handoffs, and monitoring/QA processes. Recent research from Gartner spells it out – if customer data quality and connectivity aren’t solved first, AI initiatives will fail.
How retailers are approaching adoption
Most retailers I speak with are moving in phases: internal enablement first (drive operational savings, tune agents on internal workflows, etc.), then approach gradual customer-facing rollouts. This pragmatic approach helps teams understand what “right” looks like for their brand and builds trust with stakeholders.
Common blockers are not the capabilities of the AI but data and integration: too many suppliers, siloed systems, inconsistent customer records, and legacy tech make delivering coherent, personalised agentic experiences difficult. Replacing or bridging legacy tech is a recurring priority in transformation roadmaps.
Region matters: why MENA is sprinting ahead of Europe
At the Gartner Symposium, regional differences were striking. Retailers from the MENA region (Middle East and North Africa) are noticeably further along in AI adoption than many European peers. There are three reasons for that lead:
- Less legacy: Many MENA retailers started with more modern, cloud-first architectures and fewer decades-old, fragmented systems. Without the same legacy challenges that plague many European enterprises, they can integrate data sources faster, deploy SLMs or domain-tuned models on-premises or in cloud stacks, and launch agentic experiences more rapidly.
- Stronger willingness from customers to engage with AI: Consumer sentiment in MENA tends to be more receptive to automated and agentic experiences. There’s less scepticism about interacting with bots and more tolerance for AI-driven sales and service. That cultural openness has accelerated deployment.
- Deeper pockets: Public and private capital has flowed into digital transformation across the region, and many organisations have explicit mandates to modernise customer channels. Investments prioritise rich messaging channels (WhatsApp, OTT apps, RCS) that pair naturally with agentic agents, enabling full in-channel journeys from marketing to purchase and support.
Contrast that with Europe – regulatory scrutiny is higher, legacy systems are more entrenched, and customers, overall, are more cautious about AI-driven conversations. European retailers therefore invest more in risk mitigation, compliance workflows, and internal proofs of concept before scaling outward.
Europe will certainly catch-up, but the landscape of consumer expectations and brand leadership could shift quickly in favour of those who scale successful agentic experiences now.
Practical considerations
To translate agentic promise into sustained value, retailers need to consider prioritising clear business objectives before diving into the weeds of what agentic AI can offer; ensure they have a solid data foundation; have clear guardrails and security measures designed from the beginning; plans for smooth hand-offs between human and AI agents; and a channel-first approach, understanding which channels customers prefer and which offer in-channel transactions.
Agentic AI is not a magic wand, but it is the next practical toolkit for delivering personalised, efficient and scalable retail experiences.
The immediate wins are operational savings and improved response velocity; the longer-term prize is hyper-personalisation and new commerce patterns that redefine loyalty.
My conversations in Barcelona made one thing clear: organisations across regions are ready to invest, but speed and success will be determined by data maturity, sensible risk management and the courage to roll out agentic experiences where customers are eager to use them. For many retailers – especially those in MENA – that moment has already arrived. Europe is watching closely and will follow; how quickly it closes the gap depends on solving the two classic problems of technology adoption: clear objectives and clean data.
